* I would like to give a special thanks to Òeljko Boškovie, Noam Chomsky, Guglielmo Cinque,Richard Kayne, Howard Lasnik, and David Pesetsky for reading and commenting on variousprevious versions of this manuscript. I additionally would like to thank the following people fortheir comments and questions: Sam Epstein, Kentaro Knakatani, Susumu Kuno, Alec Marantz,Jean-Yves Pollock. I also thank all the participants of the O pen Linguistics Forum at theUniversity of O ttawa titled “Challen ges of Minimalism’97” and the participants of theSpring’97 Practicum at Harvard where parts of this work were presented. I would also like to thank the editors and anonymous reviewers of this volume, especiallyGalia Alexandrova for her help and comments. Finally, I thank K rystyna and Wies �awSzczegielniak and Dorota Wojta³ for their help with Polish judgements. However, all errors aremine. 1 I will abstract here from X0 Optional Movement and will assume that John impressed Maryand Mary impressed John involve distinct numerations.

In: The Minimalist Parameter. Galia Alexandrova, Olga Arnaudova (eds.). CurrentIssues in Linguistic Theory Series. J. Benjamins.125-148 (2001)

POLISH OPTIONAL MOVEMENT *

ADAM SZCZEGIELNIAKHarvard University

1. IntroductionIn current Minimalism Move/Attract � is considered to be a Last Resort

operation driven by the need to eliminate features unreadable at the PF or LFinterfaces (see, e.g., Chomsky 1995). Crucially, if elements move because theyare forced to, then there should be no language exhibiting a structure with thesame Numeration and yet with different linear orders.1 Consequently, optionalmovement poses a problem for current Minimalism. In this paper I will discusscertain properties of various word orders in Polish and will attempt to proposea minimalist account of Polish optional movement. As a starting point, I willassume the minimalist framework proposed in Chomsky (1995, Chap.4) andlater modified in his MIT 1997 lectures.

2. Polish simple clauses and the Head Parameter In simple transitive clauses such as (1) Polish exhibits free word order:

(1) Adam zobaczy� Roberta.Adam-NOM saw Robert-ACC“Adam saw Robert.”

2 ADAM SZCZEGIELNIAK

2 I assume that Topic/Focus manifests itself by having a special stress pattern. Some of the var-iations of (1) are more natural with special Topic/Focus stress patterns.3 Boškovie & Takahashi (1995) propose that in Japanese the different word orders are createdvia Merge and the relevant features of the displaced Subje ct, Object, etc., are checked at LFthrough feature lowering which is triggered by the need to assign �-roles to arguments whichhave been optionally merged in a non-� role assigning position. I will assume that Polish OMis overt movement since it exhibits island effects and can result in feature checking.4 Interestingly, the order within the predicate, as noted by Willim (1989) is usually VO, withthe OV order usually present in certain locative constructions. This might point to an analysisof locative inversion along the lines of Collins (1997). 5 This does not imply that Polish is head-initial, since the crucial position is that of the verb.6 Willim (1995) also argues that in adjectival and nominal phrases there is a strong tendency fora complement position to follow a head and proposes that other orders are derived. Note that thenotion of Head Parameter only makes sense if we reject the proposals in Kayne (1994). I willnot discuss these issues here, since I will assume that Polish exhibits aSPECIFIERHEADCOMPLEMENT order which can be account ed for i n terms of Kayne’s(1994) LCA or the Head Parameter.

Six word orders are allowed (SVO, SOV, OSV, OVS, VSO, VOS), a ll ofwhich can be produc ed with the same non-Topic/Focus stress pattern. 2 Thevarious possible surface orders can be base generated or result from movement.The second possibility raises the question: which order is derived from which?3

In attempting to establish the basic word order of Polish, traditional grammar-ians, like Klemensiewicz (1949), assume that Polish has an SVO basic order(studies were based on cor pus data where 75% of sentences have a Subjectpreceding the predicate).4 Bartmi�ski (1973) also notes the large asymmetry inSubject and Object positions in Polish. In the majority of cases (about 80%), theSubject precedes the Object. This is regardless of whether they have morpho-logical markings distinguishing them.5 Bartmi�ski (1973:90) reports that 91%of native speakers’ judgements are that (2a) has the interpretation as in (i). How-ever, the judgements for some native speakers change when given a structure asin (2b), and the interpretation in (ii) becomes available to more speakers.

(2) a. Kurczu widzi cielu. b. Kurczu cielu widzi. chick sees calf chick calf sees (i) The chick sees a/the calf. (ii) The calf sees a/the chick.

Consequently, there seems to be a preference for an SVO order over anyother. This preference, together with the fact that Polish is a prepositional lan-guage (see 3 below), indicates that Polish is head-initial.6

(3) a. Janek poszed� do Marysi. b.*Janek poszed� Marysi do. John went to Mary John went Mary to

3POLISH OPTIONAL MOVEMENT

7 VSO orders might result from verb raising to C0. Unfortunately, for reasons of space, I willset aside the issues connected with optional X0 movement in Polish. However, I will assume thatPolish has overt raising of V to T, since Polish exhibits a rich agreement paradigm (Vikner1997). I will use the term ‘OM’ and not ‘scrambling’ since it is not clear at this stage if Polishhas scrambling as opposed to different kinds of topicalisation. However, this raises the questionwhether topicalisation can be optional. 8 The judgements are the same with factive verbs like “believe”.9 Here and elsewhere in this article SUBJ glosses ‘Subjunctive’ for occurrences of (Õe)by. (Ed.)10 I assume that wh-island effects, and probably other Subjacency effects, can be considered tobe MLC violations in current Minimalist frameworks (see Chomsky 1995, Collins 1997).

3. The domain of Polish O(ptional) M(ovement)Let us assume that Polish has optional overt synt actic movement (OM)

which is responsible for at least some of the alternations in (1).7 OM seems notto be limited to simple clauses. It can extract phrases from complements intensed subjunctive and infinitival clauses (4), but not, following Willim (1989),from tensed indicative clauses, especially with an overt matrix Subject(5).8

(4) a. [Ten dom]i Jan chce kupie ti. [this house] John wants to buy

“John wants to buy this house.”

b. [Ten dom]i pro chcu Õeby Jan obejrza� ti. [this house] want SUBJ

9 John see “I want John to see this house.”

(5) a.?[Ten dom]i pro wiem Õe Jan kupi� ti. [this house] know that John bought “I know that John bought this house.”

b.*?[Ten dom]i ja wiem Õe Jan kupi�ti. [this house] I know that John bought “I know that John bought this house.”

Willim (1989) on the basis of examples of wh-island violations like those in(7) and (8) shows that P olish OM and wh-movement are constrained bySubjacency, that is by similar locality conditions.10 However, OM is marginalin examples like (7) since, as I have shown above, OM is worse out of tensedindicative clauses (5b). However, examples in (6) show t hat when OM isallowed, a weak wh-island violation is still visible. Example (6c) shows that OMout of a subjunctive clause across a wh-word in SpecCP is worse than examples

4 ADAM SZCZEGIELNIAK

11 The relative unacceptability of (6d) is probably due to the fact that the subjunctive markerhas a tendency to encliticize to the matrix verb (see Szczegielniak 1997).12 Also, Witko³ (1993:168) argues that both OM and wh-movement produce parasitic gaps.

(i) Studentai ona przys�a�a do mnie ti nie uprzedziwszy ei o konsekwencjach. student she sent to me t not warning e about consequences“She sent me a student without warning him (me) about the consequences.”

(ii) Ona przys�a�a do mnie studentai nie uprzedziwszy ei o konsekwencjach. she sent to me student not warning e about consequences

“She sent me a student without warning him (me) about the consequences.”However, (ii) indicates that the parasitic gap argument is not very convincing since there seemsto be a gap regardless of whether a phrase has been scrambled or not (Òeljko Boškovie p.c.).13 Òeljko Boškovie also pointed out to me that the above data could still be handled by analysing

OM as derived via Merge (Boškovie & Takahashi 1995). A ban on extraction out of wh-islandsseems to be the most reliable diagnostic for movement.

without any OM (6d).11 Interestingly enough, wh-island effects disappear (in wh-movement and OM) when the wh-word is not in SpecCP, as in (6a,b).

(6) a. Janowii oni chcieli Õeby co sprzedae ti? John they wanted SUBJ what sell

“What did they want to sell John?”

b. Komui oni chcieliÕeby co sprzedae ti? whom they wanted SUBJ what sell

“Whom did they want to sell what?”

c.?? Janowii oni chcieli co Õeby sprzedae ti? John they wanted what SUBJ sell

“What did they want to sell John?”

d.?Oni chcieli co Õeby sprzedae Janowi? they wanted what SUBJ sell John “What did they want to sell John?”

(7) ?Janai oni zastanawiali siu kto odwiedzi� ti? John they wondered REFL who visited “Who did they wonder visited John?”

(8) ?Kogoi oni zastanawiali siu kto odwiedzi� ti? who they wondered REFL who visited “Who did they wonder who visited?”

The above properties have been usually analysed in terms of Move � withinthe GB frameworks.12 Hence I will assume that Polish OM is a result of move-ment.13 However, it is unclear what is its nature. Is it overt syntactic or PF (non-syntactic) movement (if one assumes that Subjacency can be a PF condition, seeHuang 1990)?

5POLISH OPTIONAL MOVEMENT

14 WCO effects arise when an operator A'-binds both a trace and a pronoun contained in anargument XP that c-commands the trace (see Lasnik & Stowell 1991, where they argue thatscrambled elements are not true operators). OM and wh-movement exhibit, however, StrongCrossover effects:

(i) *Kogoi oni lubi? (ii) *Jankai oni lubi. whoi hei likes Johni hei likes

4. Crossover properties of Polish OM and wh-movementSimilarly to Japanese and German (see Hoji & Saito 1983, Grewendorf &

Sabel 1996), Polish OM as well as clause internal wh-movement do not exhibitWeak Crossover (WCO) effects as shown in examples (9b) and (9e). However,wh-movement differs from OM in cases of long distance movement. The formerviolates WCO (9f), whereas OM does not (9c,g).14

(9) a. Jegoi przyjaciele podziwiajc Janai. his friends admire John

“Hisi friends admire Johni.”b. [Janai]k jegoi przyjaciele podziwiajc tk. John his friends admire

“Hisi friends admire Johni.”c. [Janai]k jegoi przyjaciele chcc zaprosie tk. John his friends want invite

“Hisi friends want to invite Johni.”d. [Janai]k jegoi przyjaciele chcc Õeby Robert zaprosi� tk. John his friends want SUBJ Robert invite

“Hisi friends want Robert to invite Johni.”e. [Kogoi]k jegoi przyjaciele podziwiajc tk? who his friends admire

“Whoi do hisi friends admire?” f. *[kogoi]k ty chcesz by jegoi przyjaciele podziwiali tk.

Who you want SUBJ his friends admire “Whoi I want hisi friends to admire.”

g. [Janai]k ty chcesz by jegoi przyjaciele podziwiali tk. John you want SUBJ his friends admire

“I want hisi friends to admire Johni.”This would seem to indicate that Polish OM is A-mov ement. However,

6 ADAM SZCZEGIELNIAK

15 This has been pointed out to me by Noam Chomsky (p.c.). See also Higginbotham (1983).16 Howard Lasnik (p.c.) has correctly pointed out to me that the assumption that A-movementcannot undergo recons truction is far from uncontrovers ial. Barss (1986), Belletti & Rizzi(1988), among others, have argued that A-movement also has reconstruction. However, this willnot be problematic for my approach since I will argue that reconstruction properties of OM inPolish result from the fact that it is non-feature checking movement.

WCO can also be violated in German (Webelhuth 1992, Sauerland 1996)although according to Grewendorf & Sabel (1996) there is evidence indicatingthat German scrambling is A-bar movement. Following Saito (1989, 1992), theypropose that scra mbling does not create a semantically significant operator-variable relation, hence it is not real A-bar movement (for similar proposals seeLasnik & Saito 1991). Another possibility in accounting for the above facts is toassume that jego “his” can be linked by Jan “John” but not by the wh-phrasekogo “who”.15 I will not discuss here the various possible accounts of the datain (9). For our purposes it is enough to say that the lack of WCO effects inPolish OM does not imply that OM is A-movement.

The lack of WCO effects in examples involving clause-internal movementof wh-phrases seems to pattern together with the lack of wh-island effects ofclause-internal wh-movement. I will argue that this is because wh-phrases canundergo OM clause-internally.

4.1 Restructuring and Polish OM Similarities between wh-movement and OM have led many linguists (see

Willim 1989, Witko³ 1993) to consider Polish OM to be A-bar movement. Onetypical property of such movement is its ability to undergo reconstruction, whichfollowing Chomsky 1995, I will assume that only A-bar movement can undergo.Hence, (10) is not a Condition A violation (see Saito 1992, Grewendorf & Sabel1996).16 Polish optionally moved elements also undergo reconstruction as shownin (11). (10) a. [Which pictures of himselfi]k does Rogeri hate tk ]

b. [Himselfi]k, Johni hates tk.c.[An article about herselfi]k he suggested that shei write tk tomorrow.

(11) a. [Sobiei]k Jani nie ufa tk.self John not trust

“Johni does not trust himselfi.”

7POLISH OPTIONAL MOVEMENT

17 The three Conditions of Binding Theory are as follows, following Harbert (1995:182):i. Condition A: An anaphor must be A-bound within its local domain D;ii. Condition B: A pronoun must be A-free within its local domain;iii. Condition C: An R-expression must be A-free,

where D=Tense in Polish (see Willim 1989). I w ill assume that Binding Conditions are notsubject to parametric variation (unlike D ). This in not an uncontrovers ial assumption. DavidPesetsky (p.c.) has pointed out to me that there is some evidence that Slavic languages mighthave different Binding Conditions. This would obviously be problematic for my proposals.18 See Grewendorf & Sabel (1996) for an alternative discussion on on Conditions A and B.19 Willim (1989) points out that Polish anaphors can only be Subject-bound. I will assume thatboth readings of (11c) are grammatical since the reflexive is either bound by the main clauseSubject or by PRO which is controlled by the matrix Object.20 Although I will use sometimes terminology like A vs A'-movement.21 Assuming that there is no separate AGR head.

b. [Siebiei]k on chce by Jani nienawidzi� tk. self he wants SUBJ John hate “Hei wants John to hate himselfi.”c. [Artyku� o sobiei/j]k Jani kaza� t’k Marii j PROj napisae tk. article about self John told Mary write “Jani told Maryj PROj to write an article about himselfi/herselfj.”

The fact that none of the above examples are Condition A violations, showsthat OM in cases like (11) can be undone at LF.17 Note that the optionally movedphrase does not have to necessarily reconstruct to its base position. The bindingfacts just indicate that it has to reconstruct to a position where it is c-commandedby the Subject. This seems to indicate that Polish OM behaves like A-bar or PFstyle movement.18 (11c) is interesting, since it shows that a DP optionally movedout of an infinitival clause reconstructs to its base or intermediate position, wherethe anaphor either has the main clause Subject or PRO as the antecedent.19

4.2 Possible OM landing sitesConsequently, there is evidence supporting the idea that the final landing site

of Polish OM e xhibits A-bar properties as far as Binding Conditions areconcerned. However, I will sharpen the A/A-bar di stinction and assumeChomsky’s (1995) proposals that there are two types of possible landing sites forovert movement: L-related and non-L-related.20 The former are Spec, or com-plement positions projected by lexical items. For our purposes, the set of lexicalitems is restricted to V, v and T.21 All other positions are non-L-related. I willadopt this division for Polish, however, I will show that the only relevant distin-

8 ADAM SZCZEGIELNIAK

22 Chomsky (1995:196) proposes that there might be also a broadly L-related position whichroughly corresponds to an adjoined position (see Webelhuth 1992 and Mahajan 1990).23 When T raises to C, SpecCP will be an L-related position if T features get checked.

ction for Polish OM is that between a feature checking and non-feature checkingpositions. Consequently, an element can be in SpecTP and not check any featuresof T, but has to reconstruct (to the first feature checking position), thus behavingas if it were occupying a non-L-related position.22 However, the availability ofmultiple specifiers raises the question whether all Spec positions of a givenlexical head like T ar e L-related. I will modify the definition of L-relatednessassuming that it is the function of the features that are checked in a given Specposition. Elements which check values for Case, Agreement, EPP, or are �-marked, are in L-related positions, other feature checking configurations are non-L-related.23 The adopted definition of L-relatedness predicts that if we assumethat functional heads can license non-feature-checking positions (I will argue forthis in sections 8-10), these should have properties which differentiate them fromfeature checking positions. Some properties of non-feature checking movementwill make it similar to A or A-bar movement. For example, the lack of WCOeffects would make non-f eature checking movement more similar to A-movement, on the other hand reconstruction effects would make it look like A-bar movement.24

However, we still cannot be certain as to th e nature of OM in Polish.Conceivably, it still could be a PF phenomenon, but then its similarity to A-barmovement would remain a mystery. It could be that Polish OM is essentially likeEnglish topicalisation. This would account for their similarity (compare (10)with (11)) as far as reconstruction effects are concerned. Reconstruction effects,especially if they are obligatory as in languages like Japanese (see Saito1992),could also point to the possibility that OM is sem antically vacuous, thusadditionally providing evidence that OM could be an instance of PF movement.One problem any account of OM has to address concerns the fact t hat in theminimalist model there is no place fo r overt syntactic optional movement. Ofcourse, one could resort to postulating the existence of special formal orsemantic features whose presence on various func tional heads is optional.However, such an approach runs the risk of simply restating the problem but ina more technical fashion. One plausible solution is to assume that optional wordorders are gener-ated in overt syntax but that they are a reflex of Merge, the only

9POLISH OPTIONAL MOVEMENT

operation which linguists like Fukui & Saito (1996) assume to be reprieved fromLast Resort. The following section will discuss in more detail such a proposal.

5. Polish OM and the Head ParameterFukui (1993) and Fukui & Saito (1996) assume that there is a direct relat-

ionship between the ability to undergo optional movement and the value of theHead Parameter (HP) in a given language. HP is supposed to characterise thelinear ordering of elements w ithin a clause. It allows to map c-commandrelations into linear ones. It predicts that languages should be either head-finalor head-initial. Fukui (1993) indicates that the parameter is set in a local domainand once the value is set, the linear order of the head and compleme nt isextended to all domains, even non-local ones. Fukui calls thi s a CanonicalPrecedence Relation (CPR). Hence, Polish, a prepositional language, is head-initial. Fukui (1993) and later Fukui & Saito (1996) propose that scrambling issensitive to HP. This results from the fact that there is a Parameter V(alue)P(reservation) M(easure), which following Fukui (1993:400) states:

(12) PVP Measure: A grammatical operation (Move �, in particular) thatcreates a structure that is inconsistent w ith the value of a givenparameter in a language is costly in the language, whereas one thatproduces a structure consistent with the parameter value is costless.

PVP allows us to have three kinds of movement. Last Resort movement (inthe sense of Chomsky 1995), i.e., movement which is forced for some reason(feature checking), OM which is costless and not feature checking, and, finally,costless feature-checking movement. Japanese is head-final and is correctlypredicted to have leftward OM. It follows that Polish OM cannot destroy CPR,hence it should be like English Rightward Movement, for example, extrapositiongiven below (Fukui 1993:410):

(13) a. I read a review of John’s book todayb. I read a review ti today [of John’s book]i

Fukui’s account assumes that the PP has moved rightward past the adjunct.This kind of movement does not violate the CP R and hence is cost-free.However, Polish OM can be leftward (see examples (1-7)), although it is a head-initial language. A possible account of the OM data which would be con-sistentwith the Fukui & Saito (1996) model would involve the assumption that Polishhas two kinds of OM: one feature-checking which violates HP and one which,like Japanese scrambling and Heavy NP Shift, is not feature-checking but is an

10 ADAM SZCZEGIELNIAK

25 Grewendorf & Sabel (1996:13) pose a similar question and argue, following Belletti & Rizzi(1988) that Condition A can be satisfied at any point of the derivation. I w ill not go into thedetails of this proposal, however, it does account for examples like (i) where A-moved elementsdo not seem to induce Condition A effects: (i) [Each other’si pictures ]k seem to the meni [IP tk' to be tk the most beautiful ]If this turns out to be correct, the argument (bas ed on examples like 11) that P olish OMresembles A-bar movement, as far as reconstruction is concerned, w ould be undermined.However, Noam Chomsky (p.c.) has pointed out to me that examples like (i) seem to be similarin status to constructions like: (ii) Pictures of each otheri hit the meni on theiri heads (iii) Pictures of each otheri fell on theiri headsAdditionally, it seems that an interruption with focal stress improves such examples.

instance of structure building Merge which has as its output a n ordered set ofelements and is assumed to be cost-free and not subject to Last Resort. How-ever, I will try to show in later sections that Polish OM is not feature-checking.

6. The nature of Polish OMPolish OM has Strong Crossover effects (see n.13, above), however, Willim

(1989) notes that OM can also allow elements to “escape” Condition C effects. (14) a.*Ona pokaza�a mui zdjucie Jankai z ParyÕa wczoraj.

she showed him picture John-GEN from Paris yesterday “Yesterday she showed himi John’si picture from Paris.”b. Zdjucie Jankai z ParyÕa ona pokaza�a mui wczoraj. picture John-GEN from Paris she showed him yesterday

“John’si picture from Paris she showed himi yesterday.” c.*Zdjucie Jankai z ParyÕa oni pokaza� jej wczoraj. picture John-GEN from Paris he showed her yesterday “John’si picture from Paris hei showed her yesterday.” Although example (14c) shows that reconstruction of optionally moved

phrases is obligatory (o therwise (14c) would be grammatical), the contrastbetween (14a) and (14b) is pro blematic to a framework where optionalmovement obligatorily undergoes reconstruction. Example (14b) should be justas ungrammatical as (14a) which is a Condition C violation. This, at first glance,indicates that the effects of OM must be allowed not t o be undone at the LFinterface. However, our discussion of Polish OM Condition A effects seems tocontradict the idea that optionally moved phrases do not reconstruct. 25 On thebasis of the contrast in (14) Polish OM seems to be just as problematic asGerman, Hindi or Japanese, where numerous authors have proposed that OM is

11POLISH OPTIONAL MOVEMENT

26 However, see Pesetsky (1995) for a dis cussion concerning the problems connected withdouble object constructions. 27 I want to avoid statements that Condition C applies at S-structure, since followingChomsky (1995), I assume that there is no S-structure.28 Noam Chomsky (p.c.) has also pointed out to me that the contrast in (15) could be similar tothe one between (i) and (ii): (i) She showed his picture to him. (ii) She showed him his picture. The only difference w ould be that the P olish equivalent of (i) has an empty P reposition.Hence, it might turn out that the account involving OM to an L-related position in not correct.

either A or A-bar movement or a mixture of both (see, for example, Hoji 1985,Mahajan 1990, Saito 1992, Webelhuth 1989, as well as Poole 1996).

Let us consider one possible structure of (14a) and (14b) depicted as (15a)and (15b) respectively, and see if they really involve no reconstruction.

(15) a.*[TP Ona [T pokaza�aj][vP [v tj'][VP[DP mui][V tj][DP zdjucie Jankai z ParyÕa]]]] she showed himi picture John-GEN from Paris

“She showed himi John’si picture from Paris.”b. [Zdjucie Jankai z ParyÕa]k [TP ona[T pokaza�aj][vP tk' [v tj'[VP tk [V tj][DP mui]]]] picture John’s from Paris she showed him “John’si picture from Paris she showed himi.”

In Larson’s (1988) account of double object constructions, we can see that(15a) is not the base structure from which (15b) is derived. 26 Thus, even if thedirect object were reconstructed to its base position, (15b) would still remaingrammatical. We can account for the ungrammaticality of (15a) by assuming thatthe indirect object mu “him” is optionally moved to Spec of v, where it checksCase and thus cannot reconstruct to its base position. This would entail that OMcan lead to feature checking in L-related positions.27 Another possible solutionis that mu is merged structurally higher than the direct object which does notraise above mu and thus will always be c-commanded by it. Following Koizumi(1995), we can assume that both objects check case in separate SPEC-headconfigurations and that the indirect object checks case above the direct one. Insuch an account (15a) has no movement (case checked at LF), whereas (15b) hasraising of the direct object through its case checking position to a position abovethe subject (the indirect object checks case at LF).28

Let us assume that OM always reconstructs to the last feature checking L-related position. Thus we can assume that it is a combination of two kinds ofmovement: feature checking movement to an L-related position and movementthat reconstructs which can either be topicalisation like A-bar movement or non-

12 ADAM SZCZEGIELNIAK

29 Boškovie (1997) has a similar proposal where he argues on the basis of superiority effectsthat Bulgarian objects pass through their Case checking position before raising any further. 30 CLLD constructions contain a left-dislocated object and a clitic co-indexed with it(following Cecchetto 1997). (i) Beppe, l’ho visto ieri. Beppe him-have-1SG seen yesterday

“I have seen Beppe yesterday.”31 Cecchetto propos es that after checking agreement in F 0 the clitic checks Case viaincorporation with the verb. H e also proposes that in cons tructions where there is no overttopicalised DP, there is pro in the object position which later rises to Spec of F at LF.

feature-checking movement. Both approaches will allow us to account for thereconstruction effects of (11) and for the contrast in (14). The core assumptionis that OM is a combination of two kinds of movement: A-movement which isfeature checking and non-feature checking or A-bar move ment that involvesreconstruction.

Thus in examples like (15b), I will argue that the DP object undergoes overtfeature-checking movement in order to check Case features before it raisesfurther (an thus reconstructs to that case checking position).29 This is similar toobject wh-movement in English, where, following Ura (1996), I will assume thatfunctional heads can be parametrically specified to pe rmit violations ofProcrastinate. Thus the object DP in (16) has to check Case in SpecvP before itcan raise to SpecCP (see also Chomsky 1995). (16) Whomi did you [vP ti1 [VP impress ti]]

Let me discuss a similar proposal where it is assumed that A-bar movementis always preceded by A-movement and that the former always reconstructs tothe last L-related position.

6.1 The Reconstructing Properties of CLLD Constructions and Polish OMCecchetto (1997) proposes an analysis of clitic left dislocation phenomena

(CLLD) in Romance where the topicalised DP undergoes a two-stage move-ment.30 The first stage is A-movement and is triggered by the necessity to check1-features. The next step is topicalisation which is A-bar movement. I will notdiscuss the details of Checchetto’s proposals, the crucial assumption for me isthat the XP undergoes a two-step movement and the intermediate landing site isSpec of F, where F 0 is the functional projection hosting the clitic and the 1-features which are checked by the DP and force DP-clitic agreement.31 This twostage movement analysis allows Cecchetto to predict that the topicalised DP willbehave at LF as if it were in SpecFP, since reconstruction only applies to the A-

13POLISH OPTIONAL MOVEMENT

32 The tacit assumption being that there is no downward LF movement, hence an optionallymoved object would not be able to check its Case features in Spec of v at LF. This is a reflexof the Minimal Link Condition which requires movement through all possible landing sites.This also does not preclude further reconstruction resulting from the fact that the las t featurechecking position is a non-L-related one. For example, a wh-phrase can be scrambled out of aembedded clause: (i) On my³li Õe Jan jak chcia� by Jurek za³piewa�. he thinks that John how wanted SUBJ Jurek sing

“He thinks how John wanted Jurek to sing.” The wh-word “how” jak can have scope in the most deeply embedded clause or it can havescope over the two embedded clauses, in which case the example is ambiguous (it can eitherhave the answer: “John wanted very much for Jure k to sing” or “John wanted Jurek to singloudly”).

bar part of the clitic movement. Consider the following Italian examples (fromCecchetto 1997:8-9):

(17) a. *[L’opera prima di [uno scrittore]i]k proi la scrive tk sempre (volentieri). the first work of a writeri hei it writes always with pleasureb. [L’opera prima di [uno scrittore]i]k tk la scrive sempre luii (volentieri). the first work of a writer it writes always he with pleasure “Hei writes the first work of a writeri always with pleasure.”

Example (17a) could involve reconstruction to the ar gument position t j.However, example (17b) shows that the object DP may reconstruct to a positionhigher than that of the post-verbal subject, otherwise the structur e would be aviolation of Condition C, like example (17a) with pro being co-indexed with “awriter”. Cecchetto assumes that t he position of post-verbal subjects is alwayslower than that of pre-verbal subjects. I will not discuss the problem of Italiansubject positions here (see Cardinaletti 1995 for a somewhat different approach).What is crucial for my analysis of Polish OM is that Italian topicalisation istreated as a c ombination of A and A-bar movements exhibiting A-bar recon-struction properties but only to the last A-position.

Let us propose that optionally DPs in Polish also undergo a two-st epmovement. If the second stage is non-feature check ing (violates Last Resort),then it is always optional and has to reconstruct (as far as violations ofProcrastinate I will adopt an account in Ura 1996). Of course, it could be that thesecond stage of Polish OM is in fact feature checking movement, something liketopicalisation. Regardless of the nature of this second stage of O M, followingChomsky (1995), I will assume that movement must proceed through all possiblelanding sites.32 From the above discussion, we know that the reconstruction

14 ADAM SZCZEGIELNIAK

33 Richard Kayne (p.c.) has pointed out to me that according to this account (18b) should begrammatical with both readings of the pronoun jego. I have no account for this contrast. 34 The binding judgements in (19) are the same as in (18). Howard Lasnik (p.c.) has pointed outto me that this indicates that LF Move/Attract F does not change binding relations, otherwisethere should be no contrast between (18a) (object checks case at LF as in(19a) and (18b)).

effects would remain the same. However, in section 8 I will try to provide someevidence that Polish OM is not triggered by feature checking.

Returning to the reconstruction effects of Polish OM, the examples in (18)seem to provide further evidence that scrambled elements do not behave as ifthey have undergone full reconstruction.

(18) a. Oni kaza� muj PROj przeczytae [jego i/*j ksicÕku]] he told him read his-ACC book -ACC

“He told him to read hisi/*j book.” b. [Jegoi/?j ksicÕku]k oni kaza� muj PROj przeczytae tk his-ACC book-ACC he told him read

“He told him to read his book.”

The above contrast shows that the object “his book” scram bled out of thecomplement infinitival phrase in (18b) cannot be reconstructed to its baseposition in (18a). If it were, then the scrambled pronoun should be barred frombeing co-indexed with PRO, as is the case in (18a).33 The question is what couldbe the possible intermediate feature checking site for the object DP. Examplesinvolving the Genitive of negation constructions (see Franks 1995) allow us toestablish where the Case of the embedded object is checked.

(19) a. On nie kaza� mu przeczytae jego ksicÕki. he not tell him read his-GEN book-GEN “He did not tell him to read his book.”b. Jego ksicÕki on nie kaza� mu przeczytae.

his-GEN book-GEN he not tell him read “He did not tell him to read his book.”

In (19) the matrix verb is negated and the complement of the infinitival verbin the subordinate clause changes Case from accusative to genitive. The genitiveof negation is common in Polish and it is reasonable to assume that in (19b) and(18b) the scrambled DP passes through Spec of v of the matrix verb in order tocheck case. The assumption that P olish OM reconstructs to its last featurecheck-ing position thus correctly predicts that in example s like (18b) fullreconstruction of the DP to its base position is impossible.34

15POLISH OPTIONAL MOVEMENT

35 Richards (1997) assumes that Polish multiple wh-movement patterns like Serbo-Croatian andhence is in fact an instance of multiple adjunction to IP. See also Rudin (1988).

6.2 The landing site of Polish OM Koizumi (1995) argues that the fact that multiple wh-constructions do not

violate MLC can be accounted for if we assume that multiple wh-words are inmultiple Specs of C0. Polish has multiple wh-movement, as well as multipleinstances of OM and topicalisation.35

(20) a. JANKA MARYSIA zaprosi�a. b. Janka Marysia zaprosi�a. John-EMPH Mary-EMPH invited John Mary invited “ Mary invited John.” “Mary invited John.”

(21) a. Kto kogo zabi�? b. Komu co Jan chcia� by kto da�? who whom killed to-whom what John wanted SUBJ who give “Who killed whom?” “John wanted who to give what to whom?”

Example (20a) shows that topicalisat ion can apply recursively and so canOM (20b). Example (21) shows that Polish has multiple wh-movement. In (21b)the wh-phrases raised out of the s ubordinate clause should pass through theembedded SpecCP (in order not to violate MLC which requires that movementproceeds through every possible landing site, see Chomsky 1995). Thus wewould have to assume that Polish has multiple SpecCP positions.

However, as discussed in previous sections it is not clear if the examples in(21) actually involve wh-movement to CP or if they are just instances of OM ofwh-phrases, or a combination of wh-movement and OM. However, regardless ofthe exact nature of the structures in (21), we can assume that Polish functionalheads license multiple Spec positions, thus allowing OM to move ele ments tofunctional heads like Tense, v, Comp and Focus. Note that the way I defined L-related and non-L-related positions allows me to have A-bar movement to theouter Spec of T to check some topicalisation feature, for example. This is be-cause L-relatedness is a function of the features that are checked and not of thefunctional head itself.

Before I present arguments that OM is in fact non-feature checking move-ment, let me discuss an example where elements which have undergone OM areallowed to check EPP features and thus behave like A-movement.

7. Optional movement and feature checking: The nature of the EPPWillim (1989) notes the interesting contrast below: (22a) is ungrammatical

with an overt or a null subject, (22b) is fully acceptable with a pro subject alone.

16 ADAM SZCZEGIELNIAK

36 Òeljko Boškovie (p.c.) has pointed out to me that the contrast in (22) might be a reflex of theAvoid Pronoun Principle(APP). However, the case here would involve not a general preferencefor pro over overt pronouns but wo uld only exclude overt co-indexed pronouns. Moreover,APP violations do not give rise to strong ungrammaticality judgements.

(22) a.*Oni/proi uzna� ksicÕku, którc Jani przeczyta� wczoraj za najlepszc. he/pro regarded book which John read yesterday as best “Hei regarded the book that Johni read yesterday as the best.”b. KsicÕku, którc Jani przeczyta� wczoraj, *oni/proi uzna� za najlepszc. book which John read yesterday he/pro regarded as best “The book that Johni read yesterday, hei regarded as the best.”

It seems that in (22b) the scrambled DP [ book which John read] need notreconstruct below the subject position occupied by pro.36 A natural assumptionwould be to consider the mov ement in (22b) as feature checking, likeSauerland’s (1996) proposals concerning German scrambling which he considersto be a form of topicalisation. However, if we assume that (22b) involvestopicalisation we would have to account for why OM in (22b) is disallowed.

Example (22b) shows that the scrambled DP reconstructs above pro butbelow the overt pronominal subjects. I will argue that the ungrammaticality of(22b) when it involves an overt pronominal results from the fact that the EPP inT attracts primarily phonologically overt pronominals. Consequently, the EPPcan be satisfied by optional movement of a phonologically overt DP to SpecTPinstead of having pro there. I assume that this can be formalised in terms ofAttract (as defined in Chomsky 1995), where Polish T can attract any elementwithin VP which can check the features in T. As far as EPP is concerned, anovert pronominal is preferred over a null one. However, an overt subject DP ispreferred over an overt non-subject DP (probably because the subject canautomatically check 1 and case features) and hence the contrast in judgements.This assumption predicts that the EPP is unlike any other for mal/semanticfeature, since it seems to be also sensitive to PF properties. Note that the raisedobject cannot check case or agreement features of T since it has accusative caseand does not agree with the verb. This seems to confirm the proposals in Ura(1996) where Subject properties are a function of checking 1 features in T. Thiscorrectly predicts that an optionally moved DP cannot control PRO in (23b):

(23) a. Tyi chcia�e³ [PROi zaprosie Janka] you wanted PRO invite John-ACC

17POLISH OPTIONAL MOVEMENT

37An anonymous reviewer has pointed out that (25a) should be an MLC violation. I have noaccount for its grammaticality.

b.*[Jankai]k chcia�e³ [PROi zaprosie tk] John-ACC wanted PRO invite “You wanted to invite John .”

Example (23b) also shows that the EPP is separate from case and agreementfeatures, provided one assumes that the raised DP Janka is in SpecTP. Themoved object does not agree with the verb and cannot control PRO which is atypical subject property. Neve rtheless, we can assume that its presence inSpecTP satisfies the EPP. Consequently, the preference for an overt subject DPover an object DP can be captured by the fact that subject raising checks moreT features than object raising, since the latter can only satisfy the EPP. However,it seems that the EPP cannot be satisfied by phonologically null elements if thereis a possibility of having a phonologically overt DP, hence the object DP in (22b)is in SpecTP and satisfies the EPP and is thus blocked from reconstruction. Note,however, that in examples like (23b) PRO in the subordinate clause satisfies theEPP. Furthermore, Collins (1997) proposes an analysis of locative inversion inEnglish as in (24) where the raised PP satisfies the EPP:

(24) Down the hill rolled John.

Consequently, I will assume that the optionally moved elements in Polish(like the DP object in (22b)) can satisfy EPP or check features, provided they arein the correct configuration. This is however different from proposing that OMis driven by a separate feature (see Grewendorf & Sabel 1996, Sauerland 1996).

8. The scrambling feature in PolishUp until now I have allowed the possibility that Polish OM involves non-

feature driven movement without providing any arguments for such anassumption. In this section I will argue that Polish OM does not exhibit thetypical properties of feature driven movements argued by Sauerland (1996) toindicate that German and Japanese scrambling is feature driven. His proposalsare centred around the slightly modified proposals made in Pesetsky (1982),where a distinction is proposed between Nested and Crossing paths:37

(25) a. [Jaki temat]1 ty chcesz [komu]2 PRO przedstawie t1 t2? what subject you want who present

“What subject do you want to present to who ?”

18 ADAM SZCZEGIELNIAK

38 The original examples taken by Sauerland from Pesetsky (1982) were:(i) What subject1 do you know who2 PRO to talk to t2 about t1(ii) Who1 do you know what subject2 PRO to talk to t1 about t2However, Polish has to have the prepositions pied-piped in wh-movement, hence the slight

difference in the examples . Nested and Crossing paths basically arise if one of the movedphrases dominates the other moved phrase. 39 These roughly correspond to Nested and Crossing paths, but see Sauerland (1996:4,n.2).40 All the examples are English or German structural translations from Sauerland (1996). 41 Example (29a), proposed to me by Richard Kayne (p.c.), is a modification of the originalexample in Sauerland (1996) listed in (31b), since in Polish it is possible to have optional X0

movement of the infinitival verb.

b. [Komu]2 ty chcesz [ jaki temat]1 PRO przedstawie t1 t2? who you want what subject present “To whom do you want to present what subject?”

The Polish structural counterparts of the English examples do not exhibit thesame contrast. Polish wh-movement seems to allow both Crossing (25b) andNested (25a) paths.38

Sauerland proposes a classification of syntactic movements into surfing anddiving.39 A surfi ng path is created when fi rst a superphrase together with asubphrase moves and later that subphrase is moved above the superphrase.40

(26) [Którego studenta]1 Jola prosi�a [który obrazek t1]2 by powiesie t2? which student Jola asked which picture SUBJ put up

“Which picture of which student did Jola ask to put up?”

The above kind of movement seems grammatical in Polish. Div ing Pathsinvolve the movement of the subphrase o ut of the superphrase and thesubsequent raising of the superphrase (together with the trace of the subphrase)above the extracted subphrase (example 27 is for some speakers ungrammatical-Ihave no account for this variation).

(27) [Który obrazek t1 ]2 Jola prosi�a [ którego studenta ]1 by powiesie t2? which picture Jola asked which student SUBJ put up

“Which picture did Jola ask which student to put up?”

Sauerland argues that crossing and diving paths are ill- formed if bothmovements are of the same type, the exception being crossing scrambling. Letus examine the possible combinations of both movements in Polish:41

19POLISH OPTIONAL MOVEMENT

42For some speakers of Polish (30a) is not possible without an adverbial at the end. I have noaccount for this variation.

(28) a. [Komu]1 [pytanie]2 wydawa�o siu t1 bye t2 zbyt �atwe? whom question seem REFL be too easy “To whom did the question seem too easy?”b. [Jak bardzo kochana t1 przez Jana]2 [Maria]1 wydawa�a siu bye t2? how much loved by John Mary seemed REFL be “How dearly loved by John did Mary seem to be?”

(29) a. [ Komu ]2 [ �azienku ]1 Jola t1 [wyczy³cie ti] obieca�a? who bathroom Jola clean promised

“Who did Jola promise to clean the bathroom?”b. [t1 Za pomocc szczoty wyczy³cie ]2 Jan [�azienku ]1 nieudolnie próbowa� t2. with help brush clean J. bathroom unsuccessfully tried “John tried unsuccessfully to clean the bathroom with a brush.”

(30) a. Ôe [Janowi]2 wczoraj [ Martu ]1 Adam próbowa� obiecae zabawiae t1.that John yesterday Martha Adam tried promise entertain“That Adam tried yesterday to promise John to entertain Martha.”

b. Janek [podarowa e ksicÕku t1]2 nieudolnie [Marysi]1 próbowa� t1. John give book unsuccessfully Mary tried “John tried unsuccessfully to give a book to Mary.”

Polish allows combinations of NP/Operator crossing (28a) and diving (28b)movement, it also allows a combination of scrambling/operator crossing (29a)and diving (29b) movements and, most interestingly, it allows a combination ofscrambling/scrambling crossing (30a) and divin g (30b) movements.42 Hence,Polish differs from German and Japanese in that neither German nor Japaneseallow scrambling/scrambling diving movement (see Sauerland 1996:8).

(31) a.*[Hanako-ga t1 yonda to]2 [sono-hon-o]1 Taroo-ga t2 itta. Hanako-NOM read that that book-ACC Taroo-NOM said “Taroo said that Hanako read that book.”

b.*Danny hat [t1 zu putzen]2 vergeblich [das Bad]1 t2 versuscht. Danny has to clean unsuccessfully the bathroom tried

“Danny has tried unsuccessfully to clean the bathroom.”

Sauerland (1996) accounts for the Japanese and German facts by proposing,following among others Koizumi (1995) and Müller (1996), that PMG can beaccounted for if we assume that movement is feature driven and is an instanceof Attract. He (p.8) adopts the following definitions of Attract and Cyclicity:

20 ADAM SZCZEGIELNIAK

43 See Sauerland’ s (1996) discussion on grammaticality.

i. Shortest Attract: X attracts Y, if Y is the closest phrase that can satisfya requirement of X, where Y1 is closer than Y2 to an X that c-commands

Y1 and Y2 if Y1 c-commands or dominates Y2.

ii. Cyclicity: If X is structurally lower than Y, the requirements of X mustbe satisfied prior to those of Y.

In the case of diving paths the subphrase is moved out of a superphrase (32a)which consequently raises above that subphrase.

(32) a. [XP [XP [X X] [ZP [Z Z][KP [K K]]]]] b. [XP [XP [X X] [WP [ZP [Z Z]] [WP [W W]] [KP [K K]]]

Let us assume that X attracts a scrambling feature [S]. Let us further assumethat both KP and ZP can check this feature, consequently KP should raise firstbecause of cyclicity, but ZP should raise first because of shortest Attract. Thisis why diving movement forced by the same feature is ungrammatical. Crossingmovement triggered by the same feature is a lso ungrammatical for the samereasons, the difference being that in the latter case one phrase c-commands theother (32b). The ungrammaticality of scrambling/scrambling diving movementis hence assumed to indicate that scrambling is feature driven. Consequently, thegrammaticality of (30b) compared to its Japanese and German counte rpartsseems to argue in favour of assuming that Polish OM is not feature driven.

The grammaticality of crossing paths in Polish wh-movement can beaccounted for if we assume that C 0 in Polish licenses multiple specifiers. Theembedded C0 then provides an escape hatch by attracting one wh-phrase to itsfirst specifier and the other wh-phrase to its second specifier. Note that once theinitial wh-phrase has raised to Spec1CP, it is outside the checking domain of theother wh-phrase and hence, provided C0 has multiple [+wh] to check, the otherwh-phrase can raise to the higher Spec2CP.43

9. Semantic effects of Polish OMPolish OM shows some interesting scope properties of multiply-quantified

sentences. All the examples below mean “Each boy kissed a few girls”.

21POLISH OPTIONAL MOVEMENT

44An anonymous reviewer has pointed out that t he examples in (33) have different preferredreadings depending on discourse and topic/focus stress patterns. However, what is crucial formy argument is that movement alone does not disambiguate these structures.

(33) a. KaÕdy ch�opiec poca�owa� kilka dziewczyn. (SVO) each boy kissed several girls

b. Kilka dziewczyn poca�owa� kaÕdy ch�opiec. (OVS) several girls kissed each boy c. Kilka dziewczyn kaÕdy ch�opiec poca�owa�. (OSV)

several girls each boy kissed d. Poca�owa� kaÕdy ch�opiec kilka dziewczyn. (VSO) kissed each boy several girls

e. Poca�owa� kilka dziewczyn kaÕdy ch�opiec. (VOS)kissed several girls each boy

f. KaÕdy ch�opiec kilka dziewczyn poca�owa�. (SOV) each boy several girls kissed

(33a-f) can have a reading wher e the subject or the object DP has a widescope, with the exception of (33e). (34) gives the possible corresponding LFs:

(34) a. [Every boy ]� [few girls ]�, � kissed �b. [Few girls ]� [every boy ]�, � kissed �

Example (33e) seems to prefer strongly the reading in (34a). I do not reallyhave any account of this, however, it could be that (33e) involves VP raising andnot OM of the object DP and hence the difference in judgement.

If we assume that the scope relations of multiply quantified structures areestablished at LF, then we must assume that the inability of OM to disambiguatethe above structures implies that the effects of OM at LF are undone. This initself does not imply that there is no ‘scrambling’ feature. However, togetherwith the fact that Polish allows multiple instances of OM, whe ther diving orsurfing, it seems to indicate that the lack of LF effects in Polish OM is a resultof the fact that there is no semantic or formal feature which triggers optionalmovement.44

The above discussion indicates that Polish OM is not feature driven in thesense German and Japanese OM is. Moreover, it does not induce any scopeeffects. This seems to exclude a semantic feature as the trigger for Polish OM.

10. A Tentative local economy account and conclusions

��

The account presented here assumes that Polish OM is not feature checkingin the sense of Chomsky (1995). This faces the problem why optionalmovement exists in Polish. It does not seem to be feat ure checking, hence itviolates Last Resort and should not take place.

A possible account would be to assume, following Ura (1996), that funct-ional heads are parametrically specified as to how many violations of Procrast-inate they allow during feature checking movement. In Polish, this notion wouldhave to be taken one step further: functional heads would also have to be able topermit economy violations, ev en if the movement is not feature checking.Furthermore, Polish functional heads can license Spec positions, even if theseare not feature-checking, thus allowing also optional Last Resort violations.

Let me assume that the ability to violate economy is an inherent property offunctional heads which either have this option in a given language or they don’t.This property is in fact a manifestation of the feature properties of that head. Letme assume that functional heads have a feature ! that is responsible for licensingSpecs positions. ! is a -Interpretable formal feature and it can appear in differentconfigurations with other formal/semantic features. In languages where there areno violations of Procrastinate, ! is bundled only with strong features. In caseswhere heads are parametrised to allow optional violations of Procrastinate, ! canbe bundled with weak features. A third possibility is that ! is not bundled withany feature. This I will argue is the case of Polish OM. Crucially, I will arguethat ! is not a feature which can be checked. Its existence is manifested solely bythe projection of Spec positions. Hence, the difference between Polish andGerman and Japanese OM. Note that in this system head movement which isfeature checking is a n instance of movement to a head which has to have afeature checked which is not bundled with !. The account presented h ereprovides an argument for a model where non- feature checking movement ispossible within the minimalist program. The analysis of Polish OM also showsthat there is no need to postulate the existence of broadly L-related positions. Italso predicts that non-feature checking movement should obligatorily reconstructbut only to the last feature checking position. I crucially assume that optionallymoved elements have t o check feat ures in all possible feature checkingconfigurations that they pass through. This supports the argument that movementis driven locally and that cyclicity can be derived from a local account of LastResort (see Collins 1997).

REFERENCES

��POLISH OPTIONAL MOVEMENT

Barss, Adrew. 1986. Chains and Anaphoric Dependence. Ph.D. diss., MIT, Cambridge,Mass.

Bartmi�ski, Jerzy. 1973. “Czy juzyk polski staju siu juzykiem pozycyjnym?” [Is Polishbecoming a positional language] Juzyk Polski. 2:3.81-95.Belletti, Adrianna & L uigi Rizzi. 1988. “Psych-Verbs and �-Theory”. Natural

Language and Linguistic Theory 6.291-352.Boškovie, Òeljko.1997. “On Certain Violations of the Superiority Condition, Agr0, and

Economy of Derivation”. Journal of Linguistics 33.1-28.------------ & Daiko Takahashi. 1995. “Scrambling and Last Resort”. Ms., University

of Connecticut and Tohoku University.Cardinaletti, Anna. 1995. “Subject Positions”. Ms., University of Venice.Cecchetto, Carlo. 1997. “Reconstuction in Clitic Left Dislocation”. Ms., DIPSCO,

Instituto S. Rafaele.Chomsky, Noam. 1993. “A Minimalist Program for Linguistics The ory”. The View

from Building 20: Essays in linguistics in honor of Silvain Bromberger ed. by MorrisHalle & Samuel J. Keyser. 1-52. Cambridge, Mass.: MIT Press. (Repr. as Chap.3 inChomsky 1995:167-218).

------------. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press.Collins, Chris. 1997. Local Economy. Cambridge, Mass.: MIT Press.Diesing, Molly. 1992. Indefinites. Cambridge, Mass.: MIT Press.Franks, Steven. 1995. Parameters of Slavic Morphosyntax. New York: Oxford Univ.

Press.Fukui, Naoki. 1993. “Parameters and Optionality”. Linguistic Inquiry 24.399-420.Fukui, Naoki & Mamoru Saito. 1996. “Order in the Theory of Phrase Structure and

Movement”. Frankfurt Working Papers in Linguistics. 15,Universität Frankfurt.Grewendorf, Günther & Joachim Sabel. 1996. “Multiple Specifiers and the Theory of

Adjunction: On scrambling in German and Japanese” , Frankfurt Working Papersin Linguistics. 16, Universität Frankfurt.

Harbert, Wayne. 1995. “Binding Theory, Control and pro”. Government and BindingTheory and the Minimalist Program ed. by Gert Webelhuth, 177-241. Blackwell.

Higginbotham, John. 1983. “Logical Form, Binding, and Nominals”. Linguistic Inquiry14.395-420.

Hoji, Hajime. 1985. Logical Form Constraints and Configurational Structures inJapanese. Ph.D. diss., University of Washington.

------------ & Mamoru Saito. 1983. “Weak C rossover and Move- � in Japanese”.Natural Language and Linguistic Theory 1.245-259.

Huang, C-T, Jam es. 1990. “A Not e on Rec onstruction and VP Movem ent”. Ms.,Cornell University.

Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge, Mass.: MIT Press.Klemensiewicz, Zenon. 1949. “Lokalizacja podmiotu i orzeczenia w zdaniach izolo-

�� ADAM SZCZEGIELNIAK

wanych”[Subject and predicate position in isolated sentences]. Biuletyn PTJ. 9.8-19.Koizumi, Masatoshi. 1995. Phrase Structure in Minimalist Syntax. Ph.D. diss., MIT,

Cambridge, Mass.Larson, Richard K. 1988. “On the Double Object Construction”. Linguistic Inquiry

19.335-391.Lasnik, Howard & Tim Stowell. 1991 “Weakest Crossover”. Linguistic Inquiry 22.687-

720.Mahajan, Anoop. 1990. The A/A-bar Distinction and Movement Theory. Ph.D. diss.,

MIT, Cambridge, Mass.Müller, Gereon. 1996. Incomplete Category Fronting. Habilitationsschrift, Universität

Tübingen. Pesetsky, David. 1982. Paths and Categories. Ph.D. diss., MIT, Cambridge, Mass.------------. 1995. Zero Syntax. Cambridge, Mass.: MIT Press.Poole, Geoffrey. 1996. “Optional Movement in the Minimalist Program”. Minimal

Ideas ed. by Wer ner Abraham et. al., 199-216. A msterdam & Philadelphia: J.Benjamins.

Richards, Norvin. 1997. What Moves Where When in Which Language? Ph.D. diss.,MIT.

Rudin, Catherine. 1988. “On Multiple Questions and Multiple Wh-Fronting”. NaturalLanguage and Linguistic Theory 6.445-502.

Saito, Mamoru. 1989. “Scrambling as Semantical ly Vacuous A'-movement”.Alternative Conceptions of Phrase Structure ed. by Mark Baltin & Anthony Kroch,182-200. Chicago: University of Chicago Press.

------------. 1992. “Long Distance Scrambling in Japanese”. Journal of East AsianLinguistics 1.69-118.

Sauerland, Uli. 1996. “The Interpretability of Scrambling”. Ms., MIT.Szczegielniak, Adam. 1997. “Deficient Heads and Long Head Movement in Slovak”.

Formal Approaches to Slavic Linguistics 5 ed. by Marina Lindseth & StevenFranks, 312-333. Ann Arbor: Michigan Slavic Publications.

Ura, Hiroyuki. 1996. Multiple Feature Checking: A Theory of Grammatical FunctionSplitting. Ph.D. diss., MIT, Cambridge, Mass.

Vikner, Sten. 1997. “V0-T0-I0 Movement and Inflection for Person in All Tenses”. NewComparative Syntax ed. by Liliane Haegeman. 189-213. Longman.

Webelhuth, Gert. 1989. Synatctic Saturation Phenomena and the Modern GermanicLanguages. Ph.D. diss., University of Massachusetts, Amherst.

------------. 1992. Principles and Parameters of Syntactic Saturation. New York:Oxford Univ. Press.

Willim, Ewa. 1989. On Word Order: A Government and Binding study of English andPolish. Zeszyty Naukowe Uniwersytetu Jagielo�skiego, Kraków.

------------. 1995. “The Syntax of Noun Phrases in Polish: Linearisation parameters”.

��POLISH OPTIONAL MOVEMENT

Proceedings of LP’94 ed. by Bohumil Palek. Prague: Charles Univ. Press.Witko³, Jacek. 1993. “The Theory of X-binding and Some Aspects of Scrambling in

Polish”. Papers and Studies in Contrastive Linguistics 28.153-180. Pozna�: AdamMickiewicz University.

Adam Szczegielniak

Department of LinguisticsHarvard University324 Boylston HallCambridge, MA 02138USA

Tel: (617) 495-4006e-mail: szczeg@fas.harvard.eduhttp://www.fas.harvard.edu/~szczeg